Why is the score low when valuations are at extremes? Because rich valuations tell you how far a market could fall, not when — expensive markets can stay expensive for years. The two readings are multiplied, so a near-zero on either keeps the score low: it climbs only when high valuations (loaded) meet an actual trigger (lit) — a yield-curve inversion, credit stress, a trend break or a rising recession signal. Today the market is richly valued but nothing is igniting it — loaded, not lit. Posture, not a forecast.
What's loaded, and what's not lit
The instrument multiplies a valuation severity — how stretched prices are — by whether any ignition trigger has fired, then gates the two together. Today valuation is extreme while every trigger is quiet: loaded, not lit.
| Valuation · Shiller CAPE | extreme | CAPE 41 — ~94th percentile of its own history (long-run average ~17). The “loaded” half of the model: a high multiplier on any trigger that fires. Charted below. Not itself a trigger — rich valuations can persist for years. |
| T1 · Yield-curve inversion | quiet | 3m/10y curve below zero |
| T2 · Public credit stress | quiet | Baa−10Y spread above its fire-line |
| T3 · Trend break | quiet | S&P below its 10-month average (the most reliable pre-crash signal) |
| T4 · Recession signal | quiet | rising US recession probability |
Readings as of 20 Sep 2026 · latest available monthly data, current month still in progress. Valuation severity is the Shiller CAPE's percentile versus its own history (graphed in the top panel below). Triggers read from FRED (yield curve T10Y3M, Baa−10Y credit spread, US recession probability) and S&P 500 month-end closes. The scenario also checks a forward-P/E condition (>+1.5σ) as a static entry gate — it has no separate time series. All series public-source; figures point-in-time.
The standing concern — loaded, not lit
We keep one registered drawdown scenario for US equities. It is monitored, not predicted: a structured statement of what is loaded, what would ignite it, and — crucially — what would prove it wrong.
Loaded — structural pre-conditions, all currently met
- CAPE > 35 ✓
- Buffett mktcap/GDP > 200% ✓
- fwd P/E > +1.5σ ✓
- Mag-5 concentration > 25% of S&P ✓
Not lit — ignition needs 2+ of the monitored triggers
Currently 0 of 4 tracked triggers active (see the table above). Until at least 2 fire together, the scenario stays dormant.
What would prove the concern wrong
Hard falsifier · resolves 2027-05-31. By 2027-05-31: if the S&P 500 is ≥ its entry level AND public+private credit are calm AND the 3m/10y curve is ≥ 0 AND the late-2026 to mid-2027 issuance window passed with NO ≥20% real drawdown → the bear thesis is REJECTED, R force-reset, post-mortem written.
The other side
DEBASEMENT path: record debt + managed inflation + ample liquidity (the Fed/BoJ template) can hold NOMINAL prices up for years — the bear may express as REAL (gold-denominated) erosion via the Cantillon effect, not a nominal crash. RATIONAL-BUBBLE path: AI adoption is real; while liquidity is ample and momentum holds, no nominal break. Both argue 'loaded' ≠ 'imminent'.
Known blind spots
- PRIVATE-CREDIT BLIND SPOT: T2 (public HY OAS / Baa-10Y) will read 'calm' while PIK-masked private/floating-rate credit is stressed — false comfort. Watch SOFR, 30y yield, refinancing wall, BDC discounts.
- The dated issuance wave has NO historical precedent → unbacktestable (n≈0); it is a monitored FORWARD signal, not a validated trigger.
- Severity is one analog (2000); the CI is effectively undefined (Law 3).
The instrument, in full
The four bear markets, dated
| Peak | −20% breach | Trough | Depth |
|---|---|---|---|
| Aug 2000 | Mar 2001 | Sep 2002 | -46% |
| Oct 2007 | Sep 2008 | Feb 2009 | -53% |
| Dec 2019 | Mar 2020 | Mar 2020 | -20% |
| Dec 2021 | Jun 2022 | Sep 2022 | -25% |
The four ≥20% S&P bear markets since 1996 (shaded in the chart above), dated by month: when the market peaked, first fell −20%, and bottomed. Depth = the largest fall from that peak.
Did the gauge give early warning?
| −20% breach | Curve | Credit | Trend | Recession | Gate |
|---|---|---|---|---|---|
| Mar 2001 | 7 | 11 | 18 | 7 | 6 |
| Sep 2008 | 18 | 8 | 10 | 18 | 8 |
| Mar 2020 | 9 | 0 | 17 | 6 | 1 |
| Jun 2022 | 4 |
Each cell = how many months before that bear's −20% breach the trigger first fired (blank = it never did beforehand); higher = earlier warning. The honest record: the trend break is the most consistent early signal — valuation and credit often lead by little or not at all, which is exactly why this is posture, not prediction.
Could the posture inform allocation? A modeled illustration
Hypothetical and backtested — not advice, not a live signal. A deliberately simple rule: hold the S&P, but trim equity to ~50% (rest in T-bills) only while the gate is open. Applied with as-of data (no hindsight) and trading costs, 1985–2026:
| Strategy | Return p.a. | Sharpe | Worst drawdown |
|---|---|---|---|
| Buy & hold (100% equity) | 9.4% | 0.46 | −53% |
| Gate-aware de-risk | 9.7% | 0.52 | −35% |
It roughly halves the worst drawdown (−53% → −35%) and nudges risk-adjusted return up, at negligible turnover — and the edge holds in both halves of the sample out-of-sample. The catch, stated plainly: it only cushions slow recession bears (dot-com, 2008). It did nothing for the 1987 or 2020 shocks or the 2022 rate sell-off, and structurally never will. So this is recession-bear risk reduction, not a crash shield — and emphatically not a trading signal. Past behaviour of a rule on history is not a forecast of its future.
This page reflects a monitored risk posture, not investment advice. It states a structural concern and the conditions that would confirm or falsify it; it does not predict prices or recommend trades.